Comparing crowding in human and ideal observers.

Area of Science:

• Visual perception
• Computational neuroscience
• Psychophysics

Background:

• Crowding significantly impairs object recognition when targets are near similar distractors.
• Computational models of crowding, particularly using an ideal-observer framework, are underdeveloped.
• Existing research lacks a quantitative comparison between ideal-observer models and human crowding data.

Purpose of the Study:

• To develop and validate an ideal-observer model for visual crowding.
• To compare the performance of the ideal-observer model with human performance in crowding tasks.
• To investigate the role of perceptual uncertainty in crowding.

Main Methods:

• Derived an ideal-observer model for target identification with position and identity uncertainty.
• Conducted three experiments measuring perceptual uncertainty in isolation and under crowding conditions.
• Quantitatively compared model predictions with human error rates in target identification.

Main Results:

• The ideal-observer model successfully reproduced the critical spacing characteristic of crowding, which scales with eccentricity.
• Perceptual uncertainty in isolated target identification accounted for approximately half of the errors in the crowding experiment.
• The remaining errors were explained by position and identity uncertainty increasing with flanker proximity.

Conclusions:

• Crowding can be mathematically restated as a consequence of optimal processing under uncertainty.
• Human crowding behavior may represent an adaptive strategy rather than a perceptual limitation.
• The ideal-observer model provides a robust framework for understanding crowding mechanisms.